The real-time visualization of vector maps is the most common function in CyberGIS, and it is timeconsuming, especially as the data volume becomes larger. How to improve the efficiency of visualization of large vector maps is still a significant research direction for GIS scientists. In this research, we point out that parallel optimization is appropriate for the real-time visualization of large vector maps. The main purpose of this research is to investigate a balanced decomposition approach which can balance the load of each server node in a CyberGIS cluster environment. The load balancer analyzes the spatial characteristics of the map requests and decomposes the real-time viewshed into multiple balanced sub viewsheds, so as to balance the load of all the server nodes in the cluster environment. The test results demonstrate that the approach proposed in this research has the ability to balance the load in CyberGIS cluster environment.